COMP Transmisision and Reception in LTE ADVANCED( visit for more insights)


Published on

Published in: Technology
1 Like
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

COMP Transmisision and Reception in LTE ADVANCED( visit for more insights)

  1. 1. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® LTE-ADVANCED AND 4G WIRELESS COMMUNICATIONS Coordinated Multipoint Transmission and Reception in LTE-Advanced: Deployment Scenarios and Operational Challenges Daewon Lee and Hanbyul Seo, LG Electronics Bruno Clerckx, Samsung Electronics Eric Hardouin, Orange Labs David Mazzarese, Huawei Technologies Satoshi Nagata, NTT DOCOMO Krishna Sayana, Motorola Mobility ABSTRACT and reception techniques utilize multiple trans- mit and receive antennas from multiple antenna 3GPP has completed a study on coordinated site locations, which may or may not belong to multipoint transmission and reception tech- the same physical cell, to enhance the received niques to facilitate cooperative communications signal quality as well as decrease the received across multiple transmission and reception spatial interference. This article is an overview points (e.g., cells) for the LTE-Advanced system. of potential CoMP technologies and their key In CoMP operation, multiple points coordinate deployment scenarios, which have been identi- with each other in such a way that the transmis- fied during the CoMP study performed by the sion signals from/to other points do not incur Third Generation Partnership Project (3GPP) serious interference or even can be exploited as for the Long Term Evolution (LTE)-Advanced a meaningful signal. The goal of the study is to standard development. The article also discusses evaluate the potential performance benefits of practical implementation aspects as well as oper- CoMP techniques and the implementation ational challenges for CoMP. The focus of the aspects including the complexity of the standards article is on the downlink CoMP aspects since support for CoMP. This article discusses some of uplink CoMP technologies tend to have less the deployment scenarios in which CoMP tech- standard impact; indeed, the receiver processing niques will likely be most beneficial and provides at the network side can be performed in a more an overview of CoMP schemes that might be transparent way to the UE. Note that the mech- supported in LTE-Advanced given the modern anisms discussed in this article apply to both fre- silicon/DSP technologies and backhaul designs quency-division duplex (FDD) and time-division available today. In addition, practical implemen- duplex (TDD) modes of operation. tation and operational challenges are discussed. We also assess the performance benefits of CoMP in these deployment scenarios with traffic DEPLOYMENT SCENARIOS FOR varying from low to high load. COMP TECHNOLOGY BEING INTRODUCTION CONSIDERED FOR LTE-ADVANCED The trend of increasing demand for high quality Today’s deployed LTE networks are mostly of service at the user terminal or user equipment based on macrocells only, as depicted on Fig. 1a. (UE), coupled with the shortage of wireless Such networks are said to be homogeneous spectrum, requires more advanced wireless com- because all the base stations belong to the same munication techniques to mitigate intercell inter- type and power class. In Release 8 of LTE, the ference and increase the cell edge throughput. multiple-input multiple-output (MIMO) trans- Coordinated multipoint (CoMP) transmission mission in each cell is controlled independent of 148 0163-6804/12/$25.00 © 2012 IEEE IEEE Communications Magazine • February 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  2. 2. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® that of its neighbors. Intercell interference coor- dination (ICIC) is supported by means of coor- dination message exchanges between base stations via a standardized interface named X2. The goal of ICIC is to provide the scheduler at one cell with information about the current or prospective interference situation at its neigh- bors. However, the latency, or delay with which these coordination messages can be exchanged, depends on the backhaul technology used to carry X2. This latency cannot be guaranteed to be low, as discussed in more detail later. While ICIC techniques are primarily designed for semi- static coordination, CoMP techniques target (a) (b) more dynamic coordination and may require much lower latency. Figure 1. Deployment scenarios under consideration for CoMP technology: a) While macrocells constitute the basis of a homogeneous networks; b) heterogeneous networks. mobile network’s coverage, the deployment of low-power nodes (e.g., pico- or femtocells, or relay nodes) within the macrocell is foreseen as • Coordination between the cells (sectors) a widely used solution in the future to cope with controlled by the same macro base station the ever increasing mobile traffic demands. Such (where no backhaul connection is needed) networks, called heterogeneous networks, are • Coordination between cells belonging to dif- characterized by harsh intercell interference ferent radio sites from a macro network between the macro and the low-power nodes, • Coordination between a macrocell and low- due to their closer proximity and different power power transmit/receive points within its cov- classes. Heterogeneous deployments are illus- erage, each point controlling its own cell trated in Fig. 1b. The intercell interference (with its own cell identity) effect is even more detrimental when the effec- • The same deployment as the latter, except tive cell size of the low-power node is expanded that the low-power transmit/receive points to balance the traffic load between the macro constitute distributed antennas (via RRHs) and the low-power node cell, so UE may be of the macrocell, and are thus all associated associated with a cell that does not provide the with the macrocell identity strongest received signal power [1]. Scenarios 1 and 2 are targeted for homogeneous One recent deployment trend foreseen to be scenarios (Fig. 1a), and scenarios 3 and 4 are widely applied for LTE consists of splitting the targeted for heterogeneous scenarios (Fig. 1b). base station functionalities into a baseband unit Note that different operators will have different (BBU), which performs in particular the schedul- priorities regarding the CoMP deployment sce- ing and the baseband processing, and a remote narios, leading to different implementation con- radio head (RRH), responsible for all the radio siderations. frequency (RF) operations (e.g., carrier frequen- cy transposition, filtering, power amplification). The BBU is typically placed in a technical room OVERVIEW OF (e.g., in the basement of a building) while the RRH is located close to the antenna, potentially GENERAL COMP TECHNIQUES hundreds of meters away from the BBU, both Conventionally, a set of geographically collocat- units being connected via optical fiber. In addi- ed antennas that correspond to a particular sec- tion to suppressing the feeder loss (since the torization are configured as a cell. A UE power amplifier is immediately close to the terminal is connected to a single cell at a given antenna), this approach allows BBUs managing time based on associated maximum received sig- different radio sites to be gathered in the same nal power. This cell then becomes its serving place, thereby reducing the site cost and easing cell. Given the new definitions of CoMP scenar- site maintenance. Moreover, having geographi- ios as introduced earlier, the antennas config- cally separated RRHs controlled from the same ured as a cell may not be geographically location enables either centralized BBUs jointly collocated. The term transmission point (TP) can managing the operation of several radio sites or then be used to refer to a set of collocated anten- the exchange of very low-latency coordination nas, and a cell can correspond to one or more of messages between BBUs responsible for their such TPs. Note that a single geographical site own site. RRH deployments therefore facilitate location may contain multiple TPs in case of sec- the fast coordination between transmission/ torization, with one TP corresponding to one reception points that is required for CoMP. sector. CoMP techniques can also be defined in The applicability of the CoMP functionality a more straightforward manner as the coordina- indeed depends to a great extent on the back- tion between TPs. haul characteristics (latency and capacity), which CoMP studies performed in 3GPP generally condition the type of CoMP processing that can categorized three different types of CoMP tech- be applied and the associated performance. In niques depending on the required constraints on order to account for the various possible net- the backhaul link between coordinated points work topologies and backhaul characteristics, the and the level of scheduling complexity. These study of CoMP in 3GPP has focused on the fol- types of CoMP techniques can be broadly classi- lowing scenarios: fied into coordinated scheduling and coordinat- IEEE Communications Magazine • February 2012 149C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  3. 3. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® (a) (c) (b) (d) Figure 2. General classification of CoMP techniques. ed beamforming (CS/CB), joint transmission the channel fading conditions. A dotted green (JT), and TP selection (TPS). arrow refers to an interfering cell that could CS/CB can be characterized by multiple coor- potentially be transmitting desired signals to the dinated TPs sharing only channel state informa- UE in subsequent timeframes. Hybrid methods tion (CSI) for multiple UE terminals, while data may also be implemented in a network to cope packets that need to be conveyed to a UE termi- with different types of interference. nal are available only at one TP. JT can be char- acterized by the same data transmission from multiple coordinated TPs with appropriate COORDINATED SCHEDULING AND beamforming weights. TPS can be regarded as a special form of JT, where transmission of beam- COORDINATED BEAMFORMING formed data for a given UE terminal is per- The idea of CB emerged in the mid-nineties, formed at a single TP at each time instance, mainly targeting a so-called signal-to-interfer- while the data is available at multiple coordinat- ence-plus-noise ratio (SINR) leveling problem ed TPs. [2], in which the power levels and the beamform- Figure 2 illustrates the principles of CoMP ing coefficients are calculated to achieve some CS/CB, JT, and TPS. A solid red arrow refers to common SINRs in the system or to maximize the strong interference incurred by the transmis- the minimum SINR. CS is a relatively newer sion in a neighboring cell, whereas dotted red idea. The first studies [3] mainly divide the lines represent relatively lower interference entire network in clusters and apply centralized achieved with coordination. CS/CB reduces the scheduling within each cluster in order to deter- interference level experienced by a UE terminal mine which TPs in the cluster should transmit in by appropriately selecting the beamforming each time slot and to which UE. Later studies weights of interfering points to steer the interfer- jointly use CS/CB as a tool to reduce multi-user ence toward the null space of the interfered UE and multicell interference. as represented by the dotted red arrow. JT allows Different approaches jointly combining CS one or more neighboring points to transmit the and CB have been studied in LTE-Advanced, desired signal rather than interference signals which can be classified by increasing order of from the point of view of the selected UE. TPS complexity and requirements in terms of CSI enables UE to be dynamically scheduled by the feedback and CSI sharing. most appropriate TP by exploiting changes in Coordinated beam pattern is a low feedback 150 IEEE Communications Magazine • February 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  4. 4. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® overhead approach targeting highly loaded cells deployments, a single BBU controlling multiple that consists of coordinating the precoders TPs can enable very low-latency coordination Joint transmission (beamforming matrices) in the cooperating TPs among them and allow joint scheduler imple- can be broadly in a pre-defined manner in order to reduce mentations. A joint scheduler allows resource interference variation and enable accurate link pooling, by dynamically adapting to short-term described as simulta- adaptation, while avoiding spatial CSI feedback channel conditions and different traffic loads neous transmission from the UE. Relying on a TP-specific beam experienced across the control area of the BBU. pattern in the time and/or frequency domain, the In general, large cluster sizes improve these of data to a UE TPs cycle through the fixed set of beams where gains. However, the backhaul constraints in real terminal from multi- the cycling period is decided by a central con- deployments may limit the cluster size over ple cooperating TPs. troller and conveyed to the TPs. Based on which joint transmission and scheduling may be reports of the best resources and associated performed, or it could also be constrained by the The transmission channel quality in each cycling period, the UE network to simplify scheduler operation. The could be coherent or can be scheduled in the subframes or subbands centralized scheduler can be implemented for where the coordinated beams provide the best each such cluster. non-coherent and channel conditions. Non-coherent JT may use techniques like sin- aims to improve the Precoding matrix indicator (PMI) coordina- gle-frequency network (SFN) or cyclic delay overall system tion mitigates the inter-TP interference based on diversity (CDD) schemes, which target diversity the report of a restricted or recommended PMI, gains and also enable increased transmit power throughput or similar which corresponds to a potential precoder at an to the UE. On the other hand, coherent trans- system-wide interfering TP. Assuming that the UE and base mission could be based on spatial CSI feedback stations have knowledge of a codebook com- relative to two or more TPs, which can be used performance metric. posed of quantized precoders, PMI restriction to perform MIMO transmissions from the corre- and PMI recommendation techniques constrain sponding antennas. However, better synchroniza- the interfering TPs to use only those precoders tion and much smaller timing error differences that belong to a subset of the codebook in order between transmission points are needed to real- to enhance cell edge performance. In the case of ize the full potential gains of coherent JT PMI restriction (or recommendation), the cell schemes, which may limit their applicability only edge UE calculates and reports to their serving to TPs connected by a fast backhaul. cell the restricted (or recommended) PMIs, The information theoretic bounds with ideal defined as the PMIs that create the highest (or CSI predict significant gains with multi-TP coor- lowest) interference if used as a precoder in the dination schemes [4]. However, the CSI avail- interfering TPs. With these techniques, transmit- able at the transmitter is often limited by the ters may be constrained to use a specific quan- feedback currently supported on the uplink tized PMI in the codebook as the actual channels. For 3GPP specifications, the focus has precoder in a downlink transmission. been on enabling linear precoding techniques at Interference suppression-based coordinated the transmitter, and extensions of codebook- beamforming refers to a coordinated selection of based PMI feedback that are currently support- the transmit precoders in each TP that aims at ed for single-TP MIMO are the most likely eliminating or reducing the effect of inter-TP candidates for multi-TP feedback. interference. Contrary to the PMI coordination, To enable JT, the UE may report a PMI and the base stations compute a new transmit filter corresponding channel quality indicator (CQI) based on the CSI feedback from the serving and with the assumption of JT from a set of aggre- interfering TPs, relying typically on zero-forcing gated antennas corresponding to the JT TPs. beamforming (ZFBF) or joint leakage suppres- The TPs for JT may be set up semi-statically by sion (JLS) criteria. While ZFBF filter is designed the network. On the other hand, using more TPs by forcing the interference from a reference TP incurs a network resource cost not known to the to UEs served by other TPs to zero, JLS filter is UE, since it may depend on dynamic conditions obtained by maximizing the signal-to-leakage- introduced due to traffic load, quality of service plus-noise ratio (SLNR) of the UE served by the (QoS), and fairness at the network level. More reference TP. Interference-suppression-based flexibility can be achieved if the UE could feed- CB provides more design flexibility than PMI back PMI/CQI corresponding to multiple JT coordination, but is more demanding in terms of transmission hypothesis. Clearly, there is a trade- complexity, CSI accuracy, feedback overhead, off between flexibility and overhead. One backhaul overhead, and scheduling coordination. approach is to design codebooks with a hierar- chical approach. With this approach, precoding JOINT TRANSMISSION matrix codebook may be used for single TP Joint transmission can be broadly described as a transmission to derive the PMI/CQI of each TP. simultaneous transmission of data to a UE ter- For multi-TP joint transmission, JT CQI and minal from multiple cooperating TPs. The trans- PMI can be derived by concatenating each TP’s mission could be coherent or non-coherent and CQI and PMI with inter-TP co-phasing informa- aims to improve the overall system throughput tion. More generally, multi-user JT schemes can or a similar system-wide performance metric. It also be used where the JT PMI received from is particularly helpful to improve cell edge per- multiple UE terminals can be used to perform formance by converting an interfering signal to a multi-TP ZFBF. In a typical operation, the cen- desired signal. With dense small cell deploy- tralized scheduler controlling a cluster of TPs ments and heterogeneous networks with low- obtains the PMI/CQI corresponding to joint power nodes, there could be many UE terminals transmission hypothesis from UE in the cluster that receive significant signal strength simultane- and assigns UE to TPs such that a certain metric ously from multiple TPs. Furthermore, in RRH (e.g., sum proportional rate) is maximized. IEEE Communications Magazine • February 2012 151C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  5. 5. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® TRANSMISSION POINT SELECTION (TPS) within the coverage area of the macro-cell, so In its most compact this type of network configuration is appropriate In LTE-Advanced, transmission point selection with high traffic load. But this gain comes at a form, CoMP (TPS) is investigated by extending to an orthogo- cost of spectral efficiency since data channels of transmission and nal frequency-division multiplexing (OFDM) sys- one cell collide with the reference signals or con- reception applied to tem the idea of site selection diversity trol channels of another cell. This limits the transmission power control proposed for high- applicability of CoMP techniques since JT can- the sectors of a speedHSDPA [5]. With TP selection, the signal not be applied on resource elements where such single site can just to a given UE is transmitted from a single trans- collisions occur, and CoMP gains can only be mission point within the CoMP cooperating set achieved on a limited number of resource ele- be realized by an on a certain time-frequency resource. The UE ments. upgrade of the basically reports the index of its preferred TP On the other hand, configuring the macro- scheduling and (e.g. the TP with the highest received SINR) and cell and RRHs in its coverage with the same cell corresponding CSI, which is subsequently used ID allows all the cooperating TPs to share the processing for transmission. The selected TP may dynami- same set of control channels and reference sig- algorithms since cally change from one subframe, which is the nals, by aligning the resources used for the trans- minimum signal transmit time unit equivalent to mission of data. This configuration comes at the these functions are 1ms, to another subframe via time-frequency cost of a lower control channel capacity and is already co-located. domain dynamic scheduling. In addition, if the therefore more appropriate in lightly loaded neighboring TPs remain silent by not transmit- environments. Note that the cell splitting loss ting any data, the received SINR of the UE can may be potentially recovered by designing con- be further improved. trol channels relying on user specific reference signals and transmitted only on a subset of coor- dinated TPs. Additional considerations should CONSIDERATIONS FOR THE be given to the performance of legacy terminals STANDARD AND IMPLEMENTATION which are not aware of the distributed nature of the antennas within the same cell. Some IMPLEMENTATION CONSIDERATIONS AT THE resources may need to be reserved for legacy NETWORK terminals that are compliant with earlier versions of the specifications that do not support CoMP. The deployment of CoMP techniques Finally, the network must make sure that the requires some changes in the implementation of quality of channel estimation is sufficiently accu- the network. As emphasized earlier, fiber con- rate to enable CoMP techniques. A CoMP ter- nections among distributed cooperating points minal needs not only to measure the channel and regrouping the scheduling and processing from its strongest TP, but also from TPs that are functions at a central BBU allow harnessing the farther away. The reference signals used for CSI highest CoMP gains. In its most compact form, estimation transmitted from the far-away TPs CoMP transmission and reception applied to the may be interfered by reference signals and data sectors of a single site can just be realized by an transmissions from closer TPs. In order to avoid upgrade of the scheduling and processing algo- this situation, orthogonal resources are assigned rithms since these functions are already co-locat- to reference signals transmitted by cooperating ed. Since the scheduling and processing TPs, and these resources cannot be used for data complexity increases with the number of cooper- transmissions. This loss of spectral efficiency for ating TPs, a balance needs to be found between the data is minimal given the sparse transmission the performance gains and the cost of the net- of CSI reference signals (CSI-RS) in time/fre- work upgrade. quency and is further compensated for by the Additional considerations for the deployment gains achieved in resource elements used for of CoMP in the network pertain to the propaga- CoMP transmission of the data. tion environment and the expected traffic load. The deployment of CoMP sets some con- For deployments with the same sites topology, straints on the backhaul when distant radio sites backhaul latency and capacity, and scheduling are coordinated. Regarding backhaul aspects, and processing capabilities, the configuration of two categories of radio nodes have to be distin- the TPs also has an impact on the performance guished, as introduced earlier: base stations of CoMP. In the heterogeneous scenarios 3 and (macro, pico) and RRHs. Base stations are inter- 4, where pico-cells or RRHs may be deployed in connected via the standardized X2 interface, car- hotspots or coverage holes, configuring the ried over the backhaul network, whereas RRHs cooperating TPs with the same or different phys- are connected to a BBU via point-to-point opti- ical cell identities (cell ID) offers certain trade- cal fiber. The backhaul capacity required by offs. CoMP, even for JT, depends very much on the When a macrocell and the picocells in its cov- system bandwidth, the number of UEs served in erage are configured with different cell IDs (sce- a CoMP way, the size of the cluster and the net- nario 3) with a frequency reuse factor of one, work topology (e.g. ring or star). The backhaul each cell has its own pool of resources for the latency is driven by the transport technology transmission and reception of control channels (e.g. optical fiber, microwave or copper-based and its own set of reference signals. Interference technologies), the network topology (in particu- among these channels and signals is mitigated by lar the number of routers between two nodes) scrambling sequences and by assigning orthogo- and the need to employ a security protocol over nal resources for reference signals from different the backhaul, which all taken together can lead cells. This cell splitting gain (or area splitting the latency to vary from hundreds of microsec- gain) enables a large number of connections onds to 10–20 milliseconds [6]. As a result, the 152 IEEE Communications Magazine • February 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  6. 6. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® applicability of CoMP between neighbor base tation block about the TPs transmitting to the stations depends very much on the local back- UE and the TPs that constitute interference. If One aspect that haul network deployment. In contrast, CoMP transmission hypothesis are restricted with semi- between RRHs controlled by a centralized BBU static setup by the network or by down selection needs to be consid- or co-located inter-connected BBUs seems the at the UE, complexity may be reduced with ered at the UE is most favorable deployment for CoMP across acceptable performance loss. how to measure geographically separated radio sites, because CoMP terminals should be able to measure point-to-point fiber exhibits both very low laten- an appropriate interference which fits the target- channel quality to cy and high capacity. At last, coordination ed CoMP operation in deriving CQI. Differently the TPs in the net- between the cells controlled by the same base from the conventional network where all the station remains the most straightforward CoMP neighboring cells are the source of interference, work in which CoMP deployment scenario. some coordinating cells may not cause any inter- is being operated. In addition, CoMP requires the TPs to be ference in CoMP operation by transmitting only Here, the measure- tightly time and frequency synchronized. This the CoMP terminal signal (JT) or nullifying its constraint again mostly affects CoMP between signal into the CoMP terminal direction ment can include base stations, because some additional solution (CS/CB). It is obvious that the conventional long term signal is needed to provide accurate phase synchroniza- interference measurement will lead to an overes- tion (e.g. based on Global Navigation Satellite timation of interference because the contribu- strength for tracking System), which is not needed otherwise (unless tion from the coordinated cells, which usually the user mobility and for TDD mode and multicast/broadcast service). occupies a large portion in the inter-TP interfer- the short term CSI ence, will be removed or mitigated in the actual IMPLEMENTATION CONSIDERATIONS AT THE CoMP transmission. Thus, it is required for the measurements for USER TERMINAL network to ensure that a CoMP terminal can link adaptation. measure an appropriate interference level, for One aspect that needs to be considered at the example, by measuring only the interference out- UE is how to measure the channel quality to the side of the coordinated set. Even after this con- TPs in the network in which CoMP is being sideration, residual interference may remain operated. Here, the measurement can include after CS/CB operation due to the imperfect PMI long term signal strength for tracking the user feedback, so this effect also needs to be taken mobility and the short term CSI measurements into consideration in performing the interference for link adaptation. The main purpose of the measurement. long term measurements in a conventional net- work is to prepare the handover, i.e. to report the signal strength received from each neighbor- PERFORMANCE EVALUATIONS ing cell. Regarding the CSI measurement, all the neighboring signals are treated as interference in FOR COMP calculating the suitable PMI and CQI. In con- 3GPP Radio Access Network Working Group 1 trast, in a CoMP operating network, there are has conducted a simulation campaign to evaluate multiple TPs (cells) that are involved in the com- the performance of the CoMP techniques munication to a given UE, and it is possible for described in this article. The detailed simulation a neighboring cell (from the conventional net- assumptions and simulation modeling used in work point of view) to transmit the desired sig- this campaign is described in the 3GPP Techni- nal. For a functional CoMP operation, this cal Report for LTE-Advanced CoMP [7]. The difference in the network measurement needs to evaluation results presented in this article have be considered in the UE implementation. In been performed for a specific set of assumptions addition, networks deploying advanced CoMP [8] such that they only represent one example of techniques may have multiple TPs which are the performance achievable with CoMP. CoMP associated with the same cell. In such deploy- performance results can vary depending on the ments, long term measurements may be taken assumptions on e.g. impairments modeling and between TPs belonging to the same “cell.” scheduler implementation. The long term measurements can be used to Figure 3 and Fig. 4 show the relative perfor- identify the UEs that will benefit from CoMP mance gain of example CS/CB and JT CoMP transmissions and to trigger an appropriate implementations compared to Muti-User CoMP scheme. With the long term measure- MIMO (MU-MIMO) in a non-CoMP network, ments reported by UEs, the base station is able for full-buffer and a simplified FTP traffic to identify the severity of the interference creat- model [7] respectively. MU-MIMO is a spatial ed by neighboring TPs. The CoMP measurement transmission technique where more than one set of a given UE, defined as the set of TPs user is scheduled using the same time-frequen- about which the CSI related to their link to the cy resource from a transmission point, where UE is reported, can be decided based on such co-channel separation between users is achieved measurements. by means of proper precoding and receive pro- Depending on the supported CoMP schemes, cessing. The FTP traffic model allows various a UE may be required to measure and report load factors to be modeled, and naturally leads short term feedback corresponding to the chan- to unequal loads in neighboring cells. Note that nel observed from multiple TPs. A UE is expect- performance results between homogeneous net- ed to derive PMI feedback for multiple TPs works and heterogeneous networks cannot be (either per-TP PMI or a single PMI for multiple compared due to different layout and simula- TPs) and CQI assuming multiple transmission tion parameters such as number of UEs, etc. hypotheses, where a transmission hypothesis cor- The evaluated Heterogeneous Network is based responds to an assumption in the UE link adap- on Scenario 3. For Scenario 4 the cell through- IEEE Communications Magazine • February 2012 153C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  7. 7. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® Homogeneous networks Heterogeneous networks (57 macrocells, 570 UE terminals) (57 macrocells, 4 picocells per macrocell, 1710 UE terminals) 220% 220% Cell tput (macrocell) Macrocell coverage sum tput 1111 kb/s Full buffer - relative tput gain (%) Full buffer - relative tput gain (%) 200% Cell edge UE tput 200% Cell tput Cell edge UE tput 180% 180% 160% 160% 859 kb/s 140% 140% 696 kb/s 729 kb/s 120% 120% 538 kb/s 637 kb/s 17.4 Mb/s 18.3 Mb/s 19.3 Mb/s 91.7 Mb/s 16.7 Mb/s 18.3 Mb/s 18.6 Mb/s 87.3 Mb/s 100% 100% 83.6 Mb/s 80% 80% MU-MIMO CS/CB MU-MIMO JT MU-MIMO MU-MIMO CS/CB MU-MIMO JT MU-MIMO Throughput comparison Throughput comparison Figure 3. Example homogeneous and heterogeneous networks (Scenario 3) evaluation for full buffer traffic. Homogeneous networks Heterogeneous networks (57 macrocells) (57 macrocells, 4 picocells per macrocell) 20 20 MU-MIMO Traffic model — 5% UE throughput (kb/s) MU-MIMO Traffic model - 5% UE throughput (kb/s) 18 CS/CB MU-MIMO 18 CS/CB MU-MIMO JT MU-MIMO JT MU-MIMO 16 16 14 14 12 12 10 10 8 8 6 6 4 4 2 2 0 0 0.2 0.4 0.6 0.8 1 1.2 0.2 0.4 0.6 0.8 1 1.2 Load factor (λ) Load factor (λ) Figure 4. Example homogeneous and heterogeneous networks (Scenario 3) evaluation for FTP traffic. put and the Macro-cell coverage sum through- benefit from CoMP in heterogeneous networks. put would have been identical. “Cell Tput” and JT CoMP techniques generally give larger per- “Cell-edge UE Tput” in Fig. 3 depict the aver- formance benefits than CS/CB CoMP, but at the age throughput of the entire cell and the cost of larger backhaul overheads as JT requires throughput of users in the lower 5 percent-tile TPs to share the data packet to be transmitted throughput distribution of the users in the net- to a particular UE. Note that the average cell work respectively. “Macrocell coverage sum throughput of JT CoMP in the heterogeneous Tput” in Fig. 3 depicts the sum of all cell scenario seems to be lower than for the non- throughput in the macrocell area, which CoMP case, but this is due to the fact that dif- includes picocells deployed in the macrocell. ferent scheduler algorithms were implemented The load factor in Fig. 4 is the parameter, and simulated. Indeed, due to the nature of the which is used to control the traffic load of the JT CoMP operation the same scheduler as the network; further detailed description of the single cell operations cannot be used in the sim- load factor is in [7]. ulations. However, note that the performance In addition to the results given in [9], the benefits shown by JT CoMP cannot be achieved evaluation results in this article show that CoMP by changing the scheduler parameters in single gains over the traditional single cell approach cell operations. are obtained in both homogeneous and hetero- These observations show that CoMP tech- geneous networks. CoMP gains are observed for niques are a promising technology for LTE- UEs severely affected by interference. These Advanced networks, especially where the cellular UEs are typically found at cell edges in homoge- deployment is not very regular and uniform but neous networks. In heterogeneous networks, the has rather complex cellular boundaries with numerous pico-cells create more cell boundaries, varying user distributions. CoMP will thereby and the overlay of macro and pico-cells with dif- contribute to providing to the end user a more ferent transmission powers enlarges the interfer- uniform experience of mobile broadband across ence zone. Thus more UEs become eligible to the network coverage. 154 IEEE Communications Magazine • February 2012C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®
  8. 8. C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND® CONCLUSIONS nology in South Korea. He was with LG Electronics as senior research engineer from 2006 to 2011 working on LTE and LTE Advance Physical layer technologies, and Lower cost radio In this article, we have discussed CoMP tech- 802.11ac Very High Throughput and 802.11af TV White niques and the target deployment scenarios Space Physical/MAC layer technologies. He is now a Ph.D. nodes, improved being considered as part of the LTE-Advanced candidate in Georgia Institute of Technology located in backhaul connection Atlanta, USA. radio technology standard development. Evalua- links, faster tion studies have shown that CoMP can greatly BRUNO CLERCKX ( received the ________________ improve the cell-edge user experience. Similar M.S. and Ph.D. degrees in applied science from Universite processors at the conclusions were made in the LTE Advanced catholique de Louvain. He held visiting research positions at Stanford University and EURECOM and was with Sam- base stations as well CoMP study item report, where CoMP perfor- sung Electronics from 2006 to 2011. He actively contribut- mance benefits were observed in both homoge- as user terminals, ed to 3GPP LTE/LTE-A and acted as the rapporteur for the neous and heterogeneous networks [7]. The 3GPP CoMP Study Item. He is now a Lecturer (Assistant now allow CoMP to interest for the CoMP technology is expected to Professor) at Imperial College London and an editor for IEEE Transactions on Communications. be considered grow as new network topologies (e.g., heteroge- neous networks) and geographically distributed E RIC H ARDOUIN [S’02, M’05] ( as a viable ________________ antennas for single logical cell further demand received a Ph.D. degree in signal processing and telecom- technology for solutions for interference mitigation. Lower cost munications from the University of Rennes I, France, in radio nodes, improved backhaul connection 2004. Since 2004, he has been with Orange Labs, where practical implemen- he has conducted or supervised research on physical layer links, faster processors at the base stations as and physical-MAC layer techniques for interference miti- tation and well as user terminals, now allow CoMP to be gation in mobile networks. Since 2008 he has been repre- deployment. considered as a viable technology for practical senting Orange in the physical layer standardization group of 3GPP (RAN WG1) for HSPA, LTE and LTE- implementation and deployment. Advanced. REFERENCES DAVID MAZZARESE ( graduat- ________________ ed from ENSEA, France, in 1998, and received the Ph.D. in [1] R. Madan et al., “Cell Association and Interference electrical engineering from the University of Alberta, Cana- Coordination in Heterogeneous LTE-A Cellular Net- da, in 2005. He was with Samsung Electronics from 2005 works,” IEEE JSAC, vol. 28, no. 9, Dec. 2010, pp. to 2010. He has contributed numerous technical proposals 1479–89. to IEEE 802.22, 802.22.1, 802.16m and LTE-Advanced stan- [2] F. Rashid-Farrokhi, L. Tassiulas, and K. J. R. Liu, “Joint dards. He has been with Huawei Technologies since 2010, Optimal Power Control and Beamforming in Wireless and is currently the rapporteur for the 3GPP LTE Rel-11 Networks Using Antenna Arrays,” IEEE Trans. Commun., uplink CoMP work item. vol. 46, no. 10, Oct. 1998, pp 1313–24. [3] S. Das, H. Viswanathan, and G. Rittenhouse, “Dynamic S ATOSHI N AGATA ( received the _______________ Load Balancing Through Coordinated Scheduling in B.E. and M.E. degrees from Tokyo Institute of Technology, Packet Data Systems,” IEEE Infocom, San Francisco, Tokyo, Japan, in 2001 and 2003, respectively. In 2003, he Apr. 2003. joined NTT DOCOMO, INC. He worked for the research and [4] D. Gesbert et al., “Multi-Cell MIMO Co-operative Net- development for wireless access technologies for LTE and works: A New Look at Interference,” IEEE JSAC, vol. 28, LTE-Advanced. He is currently a vice chairman of the 3GPP no. 9, Dec. 2010, pp. 1380–408. TSG RAN Working Group 1 responsible for the LTE and [5] H. Furukawa, K. Harnage, and A. Ushirokawa, “SSDT- LTE-Advanced physical layer specifications. Site Selection Diversity Transmission Power Control for CDMA Forward Link,” IEEE JSAC, vol. 18, no. 8, Aug. KRISHNA SAYANA ( received his ________________ 2000, pp. 1546–54. Ph.D. from Purdue University, USA in 2005. He was at [6] Orange, Telefónica, “R1-111174, Backhaul Modeling for Lucent CDMA technologies group during summers of CoMP,” 3GPP TSG RAN WG1, Feb. 2011. 2001, 2002 and since July 2005 has been with Motorola [7] 3GPP TR 36.819 v11.0.0, “Coordinated Multi-Point working on 4G advanced receivers and wireless access sys- Operation for LTE,” 3GPP TSG RAN WG1, Sept. 2011. tems. From 2005 to 2008, he had numerous contributions [8] LG Electronics, “R1-111628, Phase 1 CoMP Simulation to WiMAX Forum and IEEE 802.16 standards. Since 2008 Evaluation Results and Analysis for Full Buffer,” 3GPP he has been a lead delegate for Motorola in 3GPP working TSG RAN WG1, May 2011 on MIMO and CoMP technologies for LTE/LTE-A. [9] Samsung, “R1-111944, RAN1 Phase 1 CoMP Results,” 3GPP TSG RAN WG1, May 2011. H ANBYUL S EO ( received the B.S., _____________ M.E., and Ph.D. degrees in electrical engineering from BIOGRAPHIES Seoul National University, Seoul, Korea, in 2001, 2003, and 2008, respectively. He joined LG Electronics in 2008 and is D AEWON L EE ( received the B.S. and M.S. _________ currently a chief research engineer working on 3GPP LTE degree from Korea Advanced Institute of Science and Tech- and LTE-Advanced. IEEE Communications Magazine • February 2012 155C qM IEEE M ommunications q qM Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page MqM q Qmags THE WORLD’S NEWSSTAND®